Genotype-phenotype association and variant characterization in Diamond-Blackfan anemia caused by pathogenic variants in RPL35A

Abstract

Diamond Blackfan anemia (DBA) is predominantly an autosomal dominant inherited red cell aplasia primarily caused by pathogenic germline variants in ribosomal protein genes. DBA due to pathogenic RPL35A variants has been associated with large 3q29 deletions and phenotypes not common in DBA. We conducted a multi-institutional genotype-phenotype study of 45 patients with DBA associated with pathogenic RPL35A germline variants and curated the variant data on 21 additional cases from the literature. Genotype-phenotype analyses were conducted comparing patients with large deletions versus all other pathogenic variants in RPL35A. Twenty-two of the 45 cases had large deletions in RPL35A. After adjusting for multiple tests, a statistically significant association was observed between patients with a large deletion and steroid-resistant anemia, neutropenia, craniofacial abnormalities, chronic gastrointestinal problems, and intellectual disabilities (p<0.01) compared with all other pathogenic variants. Non-large deletion pathogenic variants were spread across RPL35A with no apparent hot spot and 56% of the individual family variants were observed more than once. In this, the largest known study of DBA patients with pathogenic RPL35A variants, we determined that patients with large deletions have a more severe phenotype that is clinically different from those with non-large deletion variants. Genes of interest also deleted in the 3q29 region that could be associated with some of these phenotypic features include LMLN and IQCG. Management of DBA due to large RPL35A deletions may be challenging due to complex problems and require comprehensive assessments by multiple specialists including immunologic, gastrointestinal, and developmental evaluations to provide optimal multidisciplinary care.

Introduction

Diamond Blackfan anemia (DBA) is a rare inherited bone marrow (BM) failure syndrome (IBMFS) with an estimated incidence of 5-10 per million live births.^1-5 It is characterized by failure of red blood cell (RBC) production, congenital malformations, and cancer predisposition. Classic DBA consists of profound anemia diagnosed before one year of age, macrocytosis, reticulocytopenia, and a paucity of erythroid precursors in the BM.⁶ However, DBA is a heterogenous disorder with many cases having additional symptoms or no symptoms at all. Congenital abnormalities are variable but include mostly midline craniofacial defects, renal, cardiac, and thumb abnormalities.¹Malignancies associated with DBA include myeloid neoplasia, colorectal adenocarcinoma, osteogenic sarcoma, and genitourinary cancers.^7,8 Standard treatment for anemia in DBA is long-term corticosteroids. Nearly 40% of people with DBA who respond to initial treatment become steroid-dependent, and those who fail to respond to corticosteroids require chronic RBC transfusions and iron chelation or hematopoietic cell transplant (HCT).^1,9

DBA is predominately an autosomal dominant disorder caused by pathogenic germline variants in genes encoding ribosomal proteins.¹⁰ Twenty-six ribosomal genes have been linked to DBA etiology as well as two X-linked genes, TSR2 and GATA1, which encode a ribosome chaperone and a hematopoietic transcription factor targeted by altered ribosome levels, respectively.^10,11 Although all known DBA genes are typically included in gene discovery analysis or descriptive cohort studies of DBA, genotype- phenotype studies aimed to find clinical association by gene are currently limited. Only the most frequent ribosomal genes have been examined from a genotypephenotype perspective (e.g., RPS19, RPL11, RPL5) with little to no information available for the majority of diseasecausing genes, including RPL35A.^10,12 RPL35A was first associated with DBA in 2008 and has been reported to cause about 3.5% of DBA cases.^13,14 Typically, data on DBA due to RPL35A have been confined to case reports, or as part of a larger DBA study with limited to no phenotypic information.^10,13-20

RPL35A codes for a large ribosomal subunit protein located at the telomeric end of chromosome 3q (3q29- qter); pathogenic germline variants have been reported as single-nucleotide variants (SNV), small frameshifts, inframe deletions, and large deletions involving the entire RPL35A gene with or without multiple contiguous genes deleted in the 3q29 region.^14,15 Patients with DBA caused by RPL35A have been reported to have severe anemia as well as additional phenotypes not usually shared by other DBA patients, such as immunodeficiency and autism spectrum disorder.^12,13,15 It is unclear whether these phenotypes are associated with the deletion of RPL35A itself or other genes deleted in the region. Some of these phenotypes overlap with 3q29 deletion syndrome, a clinical syndrome in which the contiguous deleted region is near, but does not include, RPL35A.¹⁵ The phenotype of 3q29 deletion syndrome is thought to be related to the genes within the deleted region and includes dysmorphic facial features, intellectual disability, musculoskeletal problems, and neuropsychiatric issues.^15,21,22 The deletion in 3q29 deletion syndrome is typically about 1.5 megabase (Mb) in size, and the consistent size is thought to be related to low copy repeat regions at each end of the deletion.^21,23,24

We conducted a multi-institutional international collaborative study of patients with DBA due to RPL35A to determine the clinical consequences of germline large deletions versus other pathogenic RPL35A variants. We also assembled variant data from additional published cases of RPL35A-associated DBA to better characterize pathogenic germline variants in this disease. The phenotypes of these patients were compared with those of the 3q29 deletion syndrome to elucidate the similarities and the differences related to variants in this region of the genome.

Methods

Study population

Patients with DBA due to RPL35A were identified within the National Cancer Institute (NCI) IBMFS cohort (clinicaltrials.gov identifier: NCT00027274),²⁵ and through collaboration with investigators from the DBA Registry of North America (DBAR: clinicaltrials. gov identifier: NCT00106015),²⁶ DBA registries from Germany, France, Italy, the Czech Republic, and Greece, and through investigators from Alabama Children’s Hospital, Arkansas Children’s Hospital, and Boston Children’s Hospital (Figure 1 and Online Supplementary Table S1). All individuals were participants in Institutional Review Board approved protocols and had signed informed consents for participating in research studies.

Additional cases of DBA for RPL35A pathogenic variant characterization were identified through a search of PubMed and review of ClinVar²⁷ and Human Gene Mutation Database (HGMD).²⁸ All data were extracted as of February 22, 2019. ClinVar variants were restricted to pathogenic or likely pathogenic RPL35A DBA variants that met the minimum requirements for data sharing and quality assurance,²⁹ and HGMD was restricted to RPL35A DBA diseasecausing mutation. Any case that was a potential duplicate of a collaborator case was excluded (Figure 1).

Clinical data extraction

Data extraction focused on clinical criteria associated with DBA and with 3q29 deletion syndrome.^22,30 A positive finding was counted as present, and a clinical finding marked absent or not stated was considered absent. Standard criteria for defining cytopenia and immunodeficiency were used, and definitions of phenotypes studied are outlined in Online Supplementary Table S2.³¹

Pathogenic variant calling and variant annotation

Pathogenic variant locations (i.e., genomic or chromosomal coordinates) were used as provided by collaborators or from publications. Methods used to identify pathogenic variants included targeted sequencing, panel testing, exome sequencing, multiplex ligation-dependent probe amplification, SNP array, or array comparative genomic hybridization (Online Supplementary Table S3). Inclusion of a variant in the study required it to be reported as pathogenic, be rare (minor allele frequency [MAF] <1% within any gnomAD ethnic subgroups³²), missense variants needed to be predicted pathogenic by meta in silico predictor programs CADD (>25)³³ and REVEL (>0.5)³⁴ (Online Supplementary Table S3), and the same case could not be included more than once. Any variants of unclear pathogenicity, including untranslated regions (UTR) and duplication, were excluded. A “large deletion” was defined as a deletion of the entire RPL35A gene. “Small frameshift” included all insertion and deletions that were not large or inframe. SnpEff was used to annotate missing chromosomal, genomic, or protein positions. ³⁵ ANNOVAR was used to annotate MAF from publicly available databases and obtain in silico prediction scores.³⁶ The St. Jude Cloud PeCan tool was used to create lollipop plots.³⁷ For large deletions, UCSC Table Browser was used to identify all deleted genes in that region.³⁸

Figure 1.Study population and types of RPL35A variants in patients with Diamond-Blackfan anemia (DBA). NCI: National Cancer Institute Inherited Bone Marrow Failure Study; DBAR: DBA registry; HGMD: Human Gene Mutation Database; Alabama Children’s: Arkansas Children’s Hospital; Boston Children’s: Boston Children’s Hospital; UTR: untranslated regions; ClinVar: ClinVar public data archive.

Genotype-phenotype statistical analysis

The cases for the genotype-phenotype analysis were made up exclusively from the ten collaborating institutions (Figure 1). The cases identified through literature review had limited to no available clinical data and were included in RPL35A variant characterization. All statistical analyses used the individual case as the analysis unit. We compared 21 different clinical features between cases with large deletions versus all other pathogenic variants using Fisher exact test or Student t-test (Stata 15 software³⁹). Bonferroni correction for 21 tests was applied. P<2.4x10^-3 were considered significant.

Results

Cohort clinical characterization and genotype-phenotype relationship

We compiled a total of 45 cases of DBA caused by RPL35A variants from ten different institutions/registries after exclusion of four cases with variants of unknown significance (Figure 1). Table 1 shows the clinical features of the 45 patients from 41 different families. Fifteen of these cases have been published previously (Online Supplementary Table S3).^{10,14,18,40-42} Nearly half of the patients (n=22) had large deletions involving the entire RPL35A along with additional genes deleted in the region. All other pathogenic variants (n=23) were either nonsense, missense, splice site, small frameshift, or inframe deletions. Six of these variants are not present in ClinVar, HGMD, or gnomAD, and are considered novel (Online Supplementary Table S3).

Forty-four percent of the cohort was male (male:female 1:1.25; P=0.14) (Table 1). Median age at identification of symptoms was at birth in patients with large deletions versus two months of age in those with other pathogenic variants (P=0.04) (Table 1). Age of last follow-up was similar for both groups with a median age of 10 and 12 years, respectively (P=0.37) (data not shown). All except two patients were alive at last follow-up. The cause of death was right ventricular outflow obstruction due to thromboembolic plaque in one patient and complications related to treatment for Wilms tumor in the other patient.

Erythrocyte adenosine deaminase (eADA) values were available only for a subset of patients. Six of 11 patients with large deletion and 5 of 13 with all other RPL35A variants had elevated eADA with no difference between the groups (P=0.682). Anemia was the most common presenting symptom and was identified in all except one patient. More than 86% of patients in both groups had severe anemia needing treatment (Table 1), but a significant number of patients with RPL35A large deletions were steroid-resistant and RBC transfusion-dependent compared with patients with other pathogenic variants in RPL35A (P=0.0023) (Table 1). The number of patients who received HCT was similar in both groups.

Neutropenia was significantly associated with large deletions (P=0.0017). In addition, a higher proportion of patients with large deletions with severe neutropenia (absolute neutrophil count <0.5x10⁹/L) required treatment for neutropenia compared with patients with other pathogenic variants (n=9 vs. n=1; P=0.004). Treatment of neutropenia included eight cases with documented use of granulocyte-colony stimulating factor (GCSF) and treatment not specified in two cases. Bone marrow findings were available only in nine patients with large deletions and three with other pathogenic variants. Patients with large deletion and severe neutropenia (n=7) appeared to have both myeloid and erythroid hypoplasia while those with no neutropenia had mainly erythroid hypoplasia (5 of 7 vs. 0 of 3; P=0.06).

Table 1.Clinical phenotypes of cases with RPL35A large deletion versus all other RPL35A pathogenic variants.

Cases with large deletions involving RPL35A were more likely to have an immune system abnormality compared with other pathogenic variants, including an immunodeficiency diagnosis (n=8 vs. n=1; P=0.01) with low immunoglobulins, low numbers of lymphocyte subsets, or poor response to childhood vaccination, and treatment with intravenous immunoglobulins. Recurrent severe infections requiring hospitalization and treatment with systemic antimicrobials were also reported more commonly in cases with large deletions (n=11 vs. n=4; P=0.03). It is noteworthy that all cases with large deletion with recurrent infections or immune system abnormalities (12 unique cases) were also neutropenic.

There were six cases with loss of function (LOF) variants (splice site=3, nonsense=1, frameshift=2) among patients with other pathogenic variants in RPL35A. Univariate analysis comparing these six cases with the 19 cases with missense or inframe deletions showed an earlier age for the diagnosis of anemia in cases with LOF variants (P=0.02) but no association with neutropenia or immunodeficiency. Only one patient (frameshift variant) had severe neutropenia that was managed with GCSF, and one (splice site defect) had neutropenia and immunodeficiency treated with immunoglobulins.

Fifty-eight percent (n=26) of the cohort had at least one congenital abnormality involving craniofacial, skeletal and/or limb, cardiac, renal, genital abnormalities, or microcephaly (Table 1). The abnormalities described in each individual case are listed in Online Supplementary Table S4. Craniofacial abnormalities were significantly more frequent in cases with RPL35A large deletions compared with other pathogenic variants (n=11 vs. n=1; P=0.0006). Cases with large deletions were also more likely to have skeletal/limb abnormalities (n=9 vs. n=2; P=0.02) as well as three or more abnormalities (n=11 vs. n=2; P=0.003). Short stature, a common feature of DBA, was identified in approximately 50% of cases in each group and there was no difference in the use of steroids in either group. Delayed development and intellectual disability including learning problems requiring an individualized education program in school were significant among cases with RPL35A large deletions compared with other pathogenic variants (n=13 vs. n=2; P=0.0004) (Table 1); one case also had a diagnosis of autism spectrum disorder. Chronic gastrointestinal (GI) problems, including chronic diarrhea, gastrostomy feeding tube-dependence, and chronic enterocolitis, were also significantly associated with large deletions compared with all other pathogenic variants (n=9 vs. n=0; P=0.0006) (Table 1).

Figure 2.Distribution of RPL35A pathogenic variants reported in patients with Diamond-Blackfan anemia (DBA). All pathogenic single nucleotide and small indel variants are represented once per family. cDNA position based on transcript NM_000996.3. Numbers and letters indicate amino acids. Dotted lines indicate exons. Gray circles: literature, HGMD, and ClinVar cases. Black circles: collaborator cases. Cluster, defined as >5 unrelated cases by family in one area, is located between codon 28 to 33 (*). The numbers within the circles indicate the number of unique families with that specific pathogenic variant.

Table 2.Pathogenic germline RPL35A variant type.

In an attempt to evaluate the phenotypes of cases with non-large deletion pathogenic variants based on their genomic location, we noted that the variants were spread across the RPL35A gene with one area of clustering from codon 28 to 33 (Figure 2). A cluster was defined as >5 unrelated cases in one area. There were nine cases from seven different families from collaborators in this cluster. The variants in six of the seven families were the same inframe deletion (c.82-84delCTT NM_000996.3). We did not find any difference in the hematologic or physical phenotype between the cases in the cluster and those spread across the RPL35A gene (data not shown).

RPL35A pathogenic variant characterization

In addition to the 45 cases obtained from collaboration with DBA registries/investigators used in the above genotype/ phenotype analysis, we identified 21 other cases of DBA due to RPL35A pathogenic variants reported in the literature We characterized the pathogenic variants in all 66 cases from 62 unrelated families. Two families had three affected family members each sharing a variant c.305 G>C or c.82_84CTT (NM_000996.3), respectively.

Twenty-eight of the 62 pathogenic RPL35A variants from unrelated families were large deletions (45%) (Table 2). Genomic co-ordinates were not available for six cases with large deletions. The size of the large deletions in 22 cases with available data ranged from 5Kb to 13.42 Mb (Online Supplementary Table S5). The number of genes deleted in these cases ranged from two to 153. Large deletions extending upstream and downstream of the 3q29 deletion syndrome region were observed in 13 cases, four cases included part but not all of the region, and five did not include the region at all (Figure 3 and Online Supplementary Table S5).

Thirty-four of the 62 pathogenic variants were non-large deletions from 19 unique collaborator families and 15 unique families from literature cases. Fifty-six percent (19 of 34) of the unique family variants were observed more than once (Online Supplementary Table S6). The inframe deletion c.82_84delCTT was the most common pathogenic variant (n=10) (Online Supplementary Table S6), followed by missense c.125A>G (n=3); three other variants were reported in two families each. All other variants were spread across the gene without clear mutational hotspots and no additional clusters were identified using the pathogenic variants from the cases in the literature (Figure 2). Most of the variants were located in the middle and 3’ end of the gene with only one pathogenic variant located near the beginning of the gene.

Discussion

We have compiled the largest cohort of patients with DBA due to pathogenic germline RPL35A variants. This international multi-institutional collaboration allowed us to combine well-curated genomic information with extensive clinical data to better characterize the genotype and DBA phenotype associated with variants involving RPL35A. We conclude that cases with RPL35A large deletions are phenotypically different from cases with other variant types in RPL35A and have a more complex clinical phenotype.

Large deletions in RPL35A were frequent, comprising 45% of the pathogenic variants in this cohort. The frequency of large deletions in published cases of DBA caused by pathogenic variants in RPL35A is unknown due to the rarity of the disease. However, our findings are consistent with what is reported in HGMD where 40% (8 of 21) of DBA cases caused by RPL35A are reported as large/gross deletions.²⁸ Fifty-six percent of non-large deletion pathogenic variants were comprised of five variants. Domain information is not well described for this gene, but we did identify apparent clustering of cases between codon 28 to 33 that may be indicative of a potentially important domain (Figure 2). A sufficient number of variants were not available for a variant cluster-phenotype analysis.

Figure 3.Cases of RPL35A large deletions ordered by size in megabases (mB). Each horizontal bar represents the starting and stopping genomic co-ordinates of a RPL35A large deletion case. For literature cases without a case identification (ID) in article, the PubMed ID (PMID) was used as case ID (Online Supplementary Table S3). Thin black vertical lines indicate genomic starting and stopping co-ordinates of 3q29 deletion syndrome; thick black vertical line is the estimated location of RPL35A. See Online Supplementary Table S5 for exact genes deleted per case.

In our genotype-phenotype analysis comparing RPL35A large deletions with all other pathogenic variants in RPL35A, patients with large deletions exhibited more severe phenotypes across multiple phenotypic domains, including severity of anemia (steroid-resistance and/or RBC transfusion-dependence), neutropenia, craniofacial abnormalities, chronic GI problems, and intellectual disabilities. This suggests that a complex phenotype may be associated with the entire RPL35A gene deletion itself or due to contiguous genes deleted in the 3q29 region. Although not well described in DBA cases caused by RPL35A variants, similar phenotypes in DBA patients with large deletions in other ribosomal genes have been reported, including multiple congenital malformations, steroid resistant anemia, and neurodevelopmental abnormalities (e.g., RPL11, RPS17, RPS19, RPS26).^18,43 Studies comparing cases with RPL35A large deletion with DBA due to large deletions involving other ribosomal protein genes may elucidate genotype-phenotype relationships and influences of contiguous gene deletions on DBA phenotypes. Notably, immunodeficiency and recurrent infections were both common features in cases with RPL35A large deletions compared with all other pathogenic variants in RPL35A (P=0.01 and 0.03, respectively). Although the number of cases is small, neutropenia appears to be contributing to immunodeficiency and recurrent infections, since all these cases were also severely neutropenic and received GCSF treatment. It is unclear from our study whether the severe neutropenia and immunodeficiency seen here is driven by RPL35A gene deletion with or without the contribution of contiguous genes deleted in the region or by other LOF variants within RPL35A itself, since severe neutropenia (accompanied by immunodeficiency in one case) was also seen in two of the six cases with other LOF variants.

Some of the clinical phenotypes of patients with DBA due to large deletions involving RPL35A are similar to those observed in 3q29 deletion syndrome, including congenital abnormalities and intellectual disability, and may be influenced by additional genes deleted in the 3q29 region.^22,23 Previous speculations suggested genes of importance in 3q29 deletion syndrome studies include PAK2, DLG1, and IL1RAP in intellectual disabilities and TP63 and OSTN in bone defects.^24,44-46 However, immunodeficiency and steroid-resistant anemia are not features found in 3q29 deletion syndrome,^22,23 making it less likely that a gene deleted in the 3q29 deletion syndrome region is the etiology of the immune and severe hematologic phenotype seen in our cohort. A case report of DBA caused by a RPL35A large deletion and immune deficiency speculated that RNF168, an autosomal recessive gene located in the 3q29 deletion syndrome region that underlies the immune deficiency Riddle Syndrome, could be a possible candidate leading to immunodeficiency in that case.¹⁵ However, only 12 cases in our cohort with available clinical information and a large deletion also had deletion of RNF168. Furthermore, cases of large deletions in our study that did not overlap with 3q29 co-ordinates (n=4) also had evidence of steroid-resistance and/or RBC transfusion-dependence (4 of 4 cases) and neutropenia requiring treatment (3 of 4 cases).

It is possible that genes in close proximity to RPL35A but not in the 3q29 deletion syndrome region may contribute to the DBA phenotypes described here. Possible candidate genes could include the closest neighbors and most frequently deleted genes LMLN and IQCG along with RPL35A. These genes were deleted in 15 and 17 of 17 large deletion cases with available co-ordinates, respectively. LMLN is a zinc-metalloprotease, and its protein product is invadolysin.^47,48 Antibodies to invadolysin have been shown to concentrate in the edge of macrophage migration, and Drosophila mutants lacking this gene show abnormal cell migration during development.^47,48 IQCG has been observed in a somatic fusion with NUP98 in T-cell acute lymphoblastic leukemia and acute myeloid leukemia. This fusion protein has been hypothesized to block differentiation of hematopoietic stem cells.^49-51 IQCG knockout zebrafish had severely impaired levels of neutrophils, monocytes, macrophages, and lymphocytes.⁴⁹ Further studies with functional data are warranted to determine the effect of other gene deletions on the phenotype characteristics of patients with large deletions which include RPL35A.

In summary, we compiled the largest cohort of DBA cases with RPL35A pathogenic variants to date. Our study is limited by small sample size, retrospective, and non-uniformity of assessment, phenotype coded as absent if not specifically mentioned in patient records, and limited family data. Since our cohort is made up of European and North American cases, the genotype-phenotype relationships may have limited generalizability. We could not examine the impact of a familial component on variation in phenotypes since we only had two families with >1 case per family. We were also limited to the 45 cases from our multi-institutional collaboration for genotype-phenotype analysis, excluding all variants found in the literature due to insufficient phenotypic data. Future studies evaluating genotype-phenotype relationships in rare diseases would benefit from detailed clinical information (as in Online Supplementary Table S4) in publications. Moreover, 30 cases included here have never been published, underscoring the importance of reporting all available cases of rare diseases to better characterize the genotype, phenotype, and potential etiology of these diseases. It is also important to include deletion analysis in molecular characterization of patients with DBA, since large deletions may not be identified on standard IBMFS gene panel testing or exome sequencing analysis.

Patients with DBA due to large deletions in RPL35A have a complex, multi-system disease phenotype with a high frequency of hematologic as well as non-hematologic problems that is clinically different from DBA associated with other pathogenic variants in RPL35A. Distinction of this subtype of DBA with RPL35A haploinsufficiency due to large deletion is important for patient management, and evaluations should include thorough investigation for immunodeficiency, GI problems, developmental delays and intellectual disabilities including neuropsychiatric problems. Identification of patients with large deletions in RPL35A should trigger early comprehensive assessment by pediatric specialists to provide optimal multidisciplinary care.

Footnotes

• Received January 7, 2020
• Accepted April 1, 2020

Correspondence

NEELAM GIRI girin@mail.nih.gov

Disclosures

DP and JV are supported by the Czech Ministry of Health Grant 16-32105A. AK is consultant of CELGENE and Novartis and and Vifor Pharma and receives Honoraria from CELGENE, Novartis and ApoPharma; CN is a consultant and is on the advisory committee for Celgene; all other authors have no conflicts of interest.

Contributions

DMG and NG wrote the manuscript and performed data analyses; NG, BPA and SAS supervised the study; all authors provided data on the patients in their studies and are responsible for the reported research and have participated in the concept and design, analysis and interpretation of data, and revising the manuscript, and approved the submitted manuscript.

Funding

Funding This research was supported by the Intramural Research Program of the Division of Cancer Epidemiology and Genetics of the National Cancer Institute; the Czech DBA registry is supported by the Czech Ministry of Health Grant 16-32105A: we acknowledge the contributions made by Ann Carr MS, CGC, Lisa Leathwood, RN, and Maureen Risch, RN provided through contract HHSN261201100018C with Westat Inc. (Rockville, MD, USA); this work utilized the computational resources of the NIH High Performance Computing Biowulf cluster.

Acknowledgments

We are indebted to the participating families, whose generosity and co-operation have made this study possible.